To recognise the emotions of the human face using CNN algorithm.
- Create an own face dataset or to import an existing dataset from the internet.
- Create a model and train it to increase the accuracy for recognize the facial emotions.
- To use the trained model for emotions recognise in real time
- Used 3 datasets - FER2013, JAFFE, Own face dataset used to train the model.
- The model has been developed with a maximum accuracy of 80.61%, with the help of python libraries
- The model is then used to recognise the emotions of a human face from the real time data using web camera
1.3.1 Fer2013 : It is a face-set uploaded by kaggle.com that contains 35887 datasets 1.3.2 JAFFE database : This was uploaded by kasrl.org that has 214 datasets 1.3.3 Our own database : We created our own database which has 360 datasets created using Viola-Jones method.
- Happy
- Sad
- Angry
- Neutral
- Scared
- Surprise
- The downloaded databases was extremely biased.
- Not easily accessible datasets.
- Creating our own data was quite difficult.
Octavio Arriaga Et Al., “Real-time Convolutional Neural Networks for Emotion and Gender Classification” ,2017.